[Feature] Add AllGather-KV sequence-parallel attention str… (#4968)

Signed-off-by: zhuhh97 <zhuhonghan@huawei.com>
Signed-off-by: dph97 <nenbaying@163.com>
Signed-off-by: zhuhh97 <zhh@unknown>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: dengyunyang <584797741@qq.com>
This commit is contained in:
authored and GitHub committed 2026-07-24 14:45:26 +00:00
1 parent 8927ec057c
commit 84bbd57e4d
33 files changed
+1181 -135

No files matched your search

+8
View File
@@ -473,6 +473,14 @@ class OmniServeCommand(CLISubcommand):
help="Ring Sequence Parallelism degree for diffusion models. "
"Equivalent to setting DiffusionParallelConfig.ring_degree.",
)
omni_config_group.add_argument(
"--allgather-degree",
dest="allgather_degree",
type=int,
default=None,
help="AllGather-KV Sequence Parallelism degree for non-causal diffusion attention. "
"Equivalent to setting DiffusionParallelConfig.allgather_degree.",
)
omni_config_group.add_argument(
"--diffusion-quantization-config",
type=json.loads,